AI system
- System
-
Delhi Police Facial Recognition System (reported by The Wire as Innefu Labs' AI Vision)
- Model
-
Not assessed
- Release
-
Exact release not established.
- Application developer
-
Innefu Labs Pvt Ltd (manufacturer, per The Wire citing the BPR&D Compendium of Equipment); supplied by Pelorus Technologies Pvt Ltd
- Role
-
The software compared facial images to propose possible matches.
- Configuration
-
The Wire identified the facial-recognition software as Innefu Labs’ AI Vision.
- Inputs and outputs
-
Not assessed
- Autonomy
-
Not assessed
- Workflow
-
A court-directed 1:many facial recognition trial on photographs of 45,000 children in children's homes, against Track Child portal data on about seven lakh missing children. Police reporting of raw match counts to the ministry, which relayed them to the court by affidavit.
- Human oversight
-
Not assessed
- Relation to AI
-
The software produced the 2,930 matches reported in the ministry's affidavit. The claim rests on the police's report of raw output, relayed by PTI, and on the vendor's blog. No cited source verifies a single match to a reunification. The police counsel described the match rate as two percent in August 2018. The Delhi High Court recorded in January 2019 that the software had not helped crack any case. The CIS India figures (10,617 matches, 3,202 identities verified) come from a report that is not among the sources.
- AI contribution
-
The software produced the 2,930 matches reported in the ministry's affidavit. The claim rests on the police's report of raw output, relayed by PTI, and on the vendor's blog. No cited source verifies a single match to a reunification. The police counsel described the match rate as two percent in August 2018. The Delhi High Court recorded in January 2019 that the software had not helped crack any case. The CIS India figures (10,617 matches, 3,202 identities verified) come from a report that is not among the sources.
Benefits
- Outcome identifiers
-
Not assessed
- Benefit
-
matched 2,930 identities on 6–10 April 2018
- Category
-
Improved Knowledge Discovery
- Status
-
Not assessed
- Beneficiary groups
-
People. Organizations. Society.
- Beneficiaries
-
Not assessed
- Reach
-
Not assessed
- Distribution and access
-
Not assessed
- Duration
-
Not assessed
- Expected?
-
Yes
- Costs and harms
-
Not assessed
Measurement
- Studies
-
Not assessed
- Study design
-
Not assessed
- Measurement
-
Not assessed
- Setting and sample
-
Not assessed
- Measures
-
matched 2,930 identities on 6–10 April 2018
- Comparison
-
Not assessed
- Results
-
matched 2,930 identities on 6–10 April 2018
- Additional value from AI
-
Not assessed
- Limitations
-
Not assessed
- Denominators
-
Not assessed
- Materials for replication
-
Not assessed
Money
- Kind of value
-
Not assessed
- Amount
-
Not assessed
- Currency and prices
-
Not assessed
- Period and unit
-
Not assessed
- Realization
-
Not assessed
- Recipients and payers
-
Not assessed
- Gross or net
-
Not assessed
- Included costs
-
Not assessed
- Comparison
-
Not assessed
- Method
-
Not assessed
- Return and payback
-
Not assessed
People and places
- Participants
-
Developer: Innefu Labs Pvt Ltd (manufacturer, per The Wire citing the BPR&D Compendium of Equipment); supplied by Pelorus Technologies Pvt Ltd. Operator:
Delhi Police (Crime Branch), using Ministry of Women and Child Development Track Child portal data. - Jurisdictions
-
Benefit: Delhi,
India. Developer: India. Operator: India. - Ownership and control
-
Not assessed
- Dependencies
-
Not assessed
- Cross-border connections
-
Not assessed
- Geopolitical implications
-
Not assessed
Follow-up
- Conditions for replication
-
Matches are verified against identity records and followed by reunification, which no cited source shows. The software's match rate is adequate. The police counsel's statement of a two percent match in August 2018 concerned the same software.
- Replication attempts
-
2 follow-up entries. Benefit did not recur: 2.
- Later findings
-
22 January 2019. On 22 January 2019 the Delhi High Court recorded in Sadhan Haldar v. NCT of Delhi that the facial recognition software had not helped crack any missing-child case, after police counsel had reported a two percent match rate in August 2018.
- Availability
-
Not assessed
- Institutional decision
-
Not assessed
- Grounds
-
Not assessed
- Action and result
-
Not assessed
- Success criterion
-
Not assessed
Evidence
- Source numbers
-
AIBD-0314-source-0001–0007
- Citations
-
Numbered references below.
- Language
-
Not assessed
- Source relationships
-
The early report relays the ministry’s matching claim. The court later recorded that no missing-child case had been cracked. The retained investigations add contrary evidence and procurement details. [2] [3] [4] [5] [6] [7]
- Original evidence
-
The early report relays the ministry’s matching claim. The court later recorded that no missing-child case had been cracked. The retained investigations add contrary evidence and procurement details. [2] [3] [4] [5] [6] [7]
- Access and versions
-
Not assessed
- Supported findings
-
Not assessed
- Evidence links
-
Not assessed
- Evidence strength
-
Not assessed
- Missing evidence
-
Not assessed
Grades
- Assessed benefits
-
People. Organizations. Society.
- Depth
-
Not assessed
- Reasons
-
People: No cited source documents a child reunited with a family as a result of the 2,930 matches. The High Court recorded on 22 January 2019 that the software had not cracked any missing-child case. Organizations: The police reported 2,930 identities “established” to the ministry. The same police told the court four months later that the software produced only a two percent match. Society: The court's own figures (5,160 children untraced of 19,916 reported in 2016–2018) are the context. No reduction attributable to the software is documented.
- Observed or intended
-
Not assessed
- Rubric
-
Not assessed
- Assessor and date
-
Not assessed
- Alternative assessments
-
Not assessed
- Fill
-
Independently reported
Entry details
- Kind
-
Not assessed
- Related entries
-
Not assessed
- Field
-
Policing / missing persons and child protection. Missing children in child-care institutions.
- Purpose
-
Procured under the Delhi High Court's direction in Sadhan Haldar to match photographs of missing and found children against the Track Child database. The April 2018 trial run was that intended use.
- Dates
-
Date basis unassessed: 2018 (precision unassessed). Updated: 25 October 2026.
- Stage
-
Not assessed
- Version
-
SHA-256: c0b5467b4ae8…
References
- Other identifiers
-
Not assessed
- Category mappings
-
Not assessed
- Brief
-
Not assessed
- Revisions
-
8 revisions. Latest: 25 October 2026.
Sources
-
News report Independent author
PTI (2018). Facial recognition system helps trace 3,000 missing children in 4 days. Business Standard, 22 April 2018.
Of them, 2,930 children could be recognised between April 6 and April 10.
-
News report Independent author Context or contrary evidence
NDTV (2018). 2 Percent Accuracy In Delhi Police Facial Recognition Software, Court Told. NDTV, 23 August 2018.
The police counsel’s two-percent match statement is contrary evidence, not proof of a benefit. PTI report, paragraphs on Rahul Mehra’s submission to the High Court.
Independent scrutiny or context, not positive corroboration.
The software is not good. There is only two per cent match,
-
Court document Independent author Context or contrary evidence
The court records that it was told that no missing-child case had been cracked. It does not find the claimed benefit. Order of 22 January 2019, ¶ 12, printed p. 6.
Independent scrutiny or context, not positive corroboration.
We are told that use of 'Facial Recognition Software' has not helped in cracking any case of missing children so far, which comes as a surprise.
-
News report Independent author Context or contrary evidence
The Wire / Pulitzer Center (2025). As AI Took Over Policing in Delhi, Who Bore the Brunt? Pulitzer Center, 2 July 2025.
The investigation supplies later scrutiny and harm accounts. Its missing-child figure is attributed to the vendor, not independently verified. The investigation’s discussion of AI Vision and the claimed identification of missing children.
Independent scrutiny or context, not positive corroboration.
During the first four days of its implementation, Innefu's product (AI Vision) helped in identifying 3000 missing children.
-
News report Independent author Context or contrary evidence
The Wire / Pulitzer Center (2025). Looking Back at the Camera: How Controversial AI Firms Shaped Delhi's Predictive Policing. Pulitzer Center, 3 July 2025.
The investigation establishes procurement and vendor relationships. Its missing-child figure repeats the developer’s account. Opening procurement paragraphs and the quotation from Innefu’s AI Vision page.
Independent scrutiny or context, not positive corroboration.
While the software was originally manufactured by Innefu Labs Pvt Ltd, it was supplied to the Delhi Police by a vendor named Pelorus Technologies Pvt. Ltd.
-
Report Independent author Context or contrary evidence
The advocacy report challenges effectiveness and legality. It must not count as an independent study supporting a positive result. Sections on the low success rate and the demand to halt facial recognition.
Independent scrutiny or context, not positive corroboration.
at least three orders of the Hon'ble High Court recording it's dissatisfaction with the facial recognition system as late as August 23, 2019
-
News report Independent author Context or contrary evidence
Manorama Yearbook (2019). Can facial recognition help fight crime? 10 July 2019.
The explainer repeats the ministry’s affidavit and the High Court’s contrary statement. It supplies no new missing-child result. Telangana model, paragraphs on Delhi’s trial and the High Court.
Independent scrutiny or context, not positive corroboration.
Of them, the police were able to identify 2,930 children with the software.